deepresearch-flow 0.5.1__py3-none-any.whl → 0.6.0__py3-none-any.whl

deepresearch_flow/paper/db.py

@@ -5,10 +5,13 @@ from __future__ import annotations
 import asyncio
 import json
 import re
+import shutil
 from pathlib import Path
 from typing import Any, Iterable
 import difflib
 
+from tqdm import tqdm
+
 import click
 import httpx
 from rich.console import Console
@@ -19,7 +22,12 @@ from deepresearch_flow.paper.config import load_config, resolve_api_keys
 from deepresearch_flow.paper.extract import parse_model_ref
 from deepresearch_flow.paper.llm import backoff_delay, call_provider
 from deepresearch_flow.paper.providers.base import ProviderError
-from deepresearch_flow.paper.template_registry import list_template_names
+from deepresearch_flow.paper.schema import SchemaError, load_schema
+from deepresearch_flow.paper.template_registry import (
+    get_stage_definitions,
+    list_template_names,
+    load_schema_for_template,
+)
 from deepresearch_flow.paper.render import resolve_render_template, render_papers
 
 try:
@@ -42,6 +50,74 @@ def write_json(path: Path, data: Any) -> None:
     path.write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")
 
 
+def load_json_payload(path: Path) -> tuple[list[dict[str, Any]], dict[str, Any] | None]:
+    try:
+        data = json.loads(path.read_text(encoding="utf-8"))
+    except json.JSONDecodeError as exc:
+        raise click.ClickException(f"Invalid JSON in {path}: {exc}") from exc
+
+    if isinstance(data, list):
+        return data, None
+    if isinstance(data, dict):
+        papers = data.get("papers")
+        if isinstance(papers, list):
+            return papers, data
+        raise click.ClickException(f"JSON object missing 'papers' list: {path}")
+
+    raise click.ClickException(f"Unsupported JSON structure in {path}")
+
+
+def is_empty_value(value: Any) -> bool:
+    if value is None:
+        return True
+    if isinstance(value, str):
+        return value.strip() == ""
+    if isinstance(value, list) or isinstance(value, dict):
+        return len(value) == 0
+    return False
+
+
+def export_compare_csv(results: list[Any], output_path: Path) -> None:
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+
+    import csv
+
+    with open(output_path, "w", newline="", encoding="utf-8") as handle:
+        writer = csv.writer(handle)
+        writer.writerow([
+            "Side", "Source Hash", "Title", "Match Status", "Match Type",
+            "Match Score", "Source Path", "Other Source Hash", "Other Title",
+            "Other Source Path", "Lang"
+        ])
+        for result in results:
+            writer.writerow([
+                result.side,
+                result.source_hash,
+                result.title,
+                result.match_status,
+                result.match_type or "",
+                f"{result.match_score:.4f}",
+                result.source_path or "",
+                result.other_source_hash or "",
+                result.other_title or "",
+                result.other_source_path or "",
+                result.lang or "",
+            ])
+
+
+def export_only_in_b_paths(results: list[Any], output_path: Path) -> int:
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    lines = []
+    for result in results:
+        if result.side != "B" or result.match_status != "only_in_B":
+            continue
+        if result.source_path:
+            lines.append(result.source_path)
+
+    output_path.write_text("\n".join(lines) + ("\n" if lines else ""), encoding="utf-8")
+    return len(lines)
+
+
 def normalize_authors(value: Any) -> list[str]:
     if value is None:
         return []
@@ -133,6 +209,18 @@ def parse_year_month(date_str: str | None) -> tuple[str | None, str | None]:
     return year, None
 
 
+def resolve_relative_path(path: Path, roots: Iterable[Path]) -> Path:
+    resolved = path.resolve()
+    roots_by_depth = sorted(roots, key=lambda r: len(str(r.resolve())), reverse=True)
+    for root in roots_by_depth:
+        root_resolved = root.resolve()
+        try:
+            return resolved.relative_to(root_resolved)
+        except ValueError:
+            continue
+    return Path(path.name)
+
+
 def clean_journal_name(name: str | None) -> str:
     if not name:
         return "Unknown"
@@ -266,6 +354,147 @@ def parse_tag_list(text: str) -> list[str]:
 
 
 def register_db_commands(db_group: click.Group) -> None:
+    @db_group.group("snapshot")
+    def snapshot_group() -> None:
+        """Build production snapshot artifacts (SQLite + static export)."""
+
+    @snapshot_group.command("build")
+    @click.option("-i", "--input", "input_paths", multiple=True, required=True, help="Input JSON file path")
+    @click.option("-b", "--bibtex", "bibtex_path", default=None, help="Optional BibTeX file path")
+    @click.option(
+        "--md-root",
+        "md_roots",
+        multiple=True,
+        default=(),
+        help="Optional markdown root directory (repeatable) for source viewing",
+    )
+    @click.option(
+        "--md-translated-root",
+        "md_translated_roots",
+        multiple=True,
+        default=(),
+        help="Optional markdown root directory (repeatable) for translated viewing",
+    )
+    @click.option(
+        "--pdf-root",
+        "pdf_roots",
+        multiple=True,
+        default=(),
+        help="Optional PDF root directory (repeatable) for PDF discovery",
+    )
+    @click.option("--output-db", "output_db", default="paper_snapshot.db", show_default=True, help="Output DB path")
+    @click.option(
+        "--static-export-dir",
+        "static_export_dir",
+        default="paper-static",
+        show_default=True,
+        help="Output directory for hashed static assets",
+    )
+    @click.option(
+        "--previous-snapshot-db",
+        "previous_snapshot_db",
+        default=None,
+        help="Optional previous snapshot DB path for identity continuity",
+    )
+    def snapshot_build(
+        input_paths: tuple[str, ...],
+        bibtex_path: str | None,
+        md_roots: tuple[str, ...],
+        md_translated_roots: tuple[str, ...],
+        pdf_roots: tuple[str, ...],
+        output_db: str,
+        static_export_dir: str,
+        previous_snapshot_db: str | None,
+    ) -> None:
+        """Build a production snapshot (SQLite + static export)."""
+        from deepresearch_flow.paper.snapshot.builder import SnapshotBuildOptions, build_snapshot
+
+        opts = SnapshotBuildOptions(
+            input_paths=[Path(path) for path in input_paths],
+            bibtex_path=Path(bibtex_path) if bibtex_path else None,
+            md_roots=[Path(root) for root in md_roots],
+            md_translated_roots=[Path(root) for root in md_translated_roots],
+            pdf_roots=[Path(root) for root in pdf_roots],
+            output_db=Path(output_db),
+            static_export_dir=Path(static_export_dir),
+            previous_snapshot_db=Path(previous_snapshot_db) if previous_snapshot_db else None,
+        )
+        build_snapshot(opts)
+        click.echo(f"Wrote snapshot DB: {opts.output_db}")
+        click.echo(f"Wrote static export: {opts.static_export_dir}")
+
+    @db_group.group("api")
+    def api_group() -> None:
+        """Read-only JSON API server backed by a snapshot DB."""
+
+    @api_group.command("serve")
+    @click.option("--snapshot-db", "snapshot_db", required=True, help="Path to paper_snapshot.db")
+    @click.option(
+        "--static-base-url",
+        "static_base_url",
+        default=None,
+        help="Static asset base URL (e.g. https://static.example.com)",
+    )
+    @click.option(
+        "--cors-origin",
+        "cors_origins",
+        multiple=True,
+        default=(),
+        help="Allowed CORS origin (repeatable; default is '*')",
+    )
+    @click.option("--max-query-length", "max_query_length", type=int, default=500, show_default=True)
+    @click.option("--max-page-size", "max_page_size", type=int, default=100, show_default=True)
+    @click.option("--max-pagination-offset", "max_pagination_offset", type=int, default=10000, show_default=True)
+    @click.option("--host", default="127.0.0.1", show_default=True, help="Bind host")
+    @click.option("--port", default=8001, type=int, show_default=True, help="Bind port")
+    def api_serve(
+        snapshot_db: str,
+        static_base_url: str | None,
+        cors_origins: tuple[str, ...],
+        max_query_length: int,
+        max_page_size: int,
+        max_pagination_offset: int,
+        host: str,
+        port: int,
+    ) -> None:
+        """Serve the snapshot-backed JSON API."""
+        import os
+        import uvicorn
+
+        from deepresearch_flow.paper.snapshot.api import ApiLimits, create_app
+
+        static_base_url_value = (
+            static_base_url
+            or os.getenv("PAPER_DB_STATIC_BASE")
+            or os.getenv("PAPER_DB_STATIC_BASE_URL")
+            or ""
+        )
+        api_base_url = os.getenv("PAPER_DB_API_BASE") or ""
+        if api_base_url and host == "127.0.0.1" and port == 8001:
+            from urllib.parse import urlparse
+
+            parsed = urlparse(api_base_url)
+            if not parsed.scheme:
+                parsed = urlparse(f"http://{api_base_url}")
+            if parsed.hostname:
+                host = parsed.hostname
+            if parsed.port:
+                port = parsed.port
+        cors_allowed = list(cors_origins) if cors_origins else ["*"]
+        limits = ApiLimits(
+            max_query_length=max_query_length,
+            max_page_size=max_page_size,
+            max_pagination_offset=max_pagination_offset,
+        )
+        app = create_app(
+            snapshot_db=Path(snapshot_db),
+            static_base_url=static_base_url_value,
+            cors_allowed_origins=cors_allowed,
+            limits=limits,
+        )
+        click.echo(f"Serving API on http://{host}:{port} (Ctrl+C to stop)")
+        uvicorn.run(app, host=host, port=port, log_level="info")
+
     @db_group.command("append-bibtex")
     @click.option("-i", "--input", "input_path", required=True, help="Input JSON file path")
     @click.option("-b", "--bibtex", "bibtex_path", required=True, help="Input BibTeX file path")
@@ -782,15 +1011,195 @@ def register_db_commands(db_group: click.Group) -> None:
         write_json(Path(output_path), filtered)
         click.echo(f"Filtered down to {len(filtered)} papers")
 
-    @db_group.command("merge")
+    @db_group.group("merge")
+    def merge_group() -> None:
+        """Merge paper JSON inputs."""
+
+    def _summarize_merge(output_path: Path, merged: list[dict[str, Any]], *, input_count: int) -> None:
+        field_set: set[str] = set()
+        for item in merged:
+            if isinstance(item, dict):
+                field_set.update(item.keys())
+        field_list = sorted(field_set)
+
+        console = Console()
+        summary = Table(title="Merge Summary")
+        summary.add_column("Metric", style="bold")
+        summary.add_column("Value")
+        summary.add_row("Inputs", str(input_count))
+        summary.add_row("Items", str(len(merged)))
+        summary.add_row("Fields", str(len(field_list)))
+        summary.add_row("Output", str(output_path))
+        console.print(summary)
+
+        if field_list:
+            field_table = Table(title="Fields")
+            field_table.add_column("Name")
+            for name in field_list:
+                field_table.add_row(name)
+            console.print(field_table)
+
+    @merge_group.command("library")
+    @click.option("-i", "--inputs", "input_paths", multiple=True, required=True, help="Input JSON files")
+    @click.option("-o", "--output", "output_path", required=True, help="Output JSON file path")
+    def merge_library(input_paths: Iterable[str], output_path: str) -> None:
+        paths = [Path(path) for path in input_paths]
+        merged: list[dict[str, Any]] = []
+        for path in paths:
+            merged.extend(load_json(path))
+        output = Path(output_path)
+        write_json(output, merged)
+        _summarize_merge(output, merged, input_count=len(paths))
+
+    @merge_group.command("templates")
     @click.option("-i", "--inputs", "input_paths", multiple=True, required=True, help="Input JSON files")
     @click.option("-o", "--output", "output_path", required=True, help="Output JSON file path")
-    def merge_papers(input_paths: Iterable[str], output_path: str) -> None:
+    def merge_templates(input_paths: Iterable[str], output_path: str) -> None:
+        from deepresearch_flow.paper import db_ops
+
+        paths = [Path(path) for path in input_paths]
+        inputs = db_ops._load_paper_inputs(paths)
+        if not inputs:
+            raise click.ClickException("No input JSON files provided")
+
+        groups: list[dict[str, Any]] = []
+        base_papers: list[dict[str, Any]] = []
+        hash_to_group: dict[str, int] = {}
+        paper_id_to_group: dict[int, int] = {}
+        paper_index: dict[str, list[dict[str, Any]]] = {}
+
+        def rebuild_index() -> None:
+            nonlocal paper_index, paper_id_to_group
+            paper_index = db_ops._build_paper_index(base_papers)
+            paper_id_to_group = {id(paper): idx for idx, paper in enumerate(base_papers)}
+
+        def add_group(template_tag: str, paper: dict[str, Any]) -> None:
+            group = {
+                "templates": {template_tag: paper},
+                "template_order": [template_tag],
+            }
+            groups.append(group)
+            base_papers.append(paper)
+            source_hash = str(paper.get("source_hash") or "")
+            if source_hash:
+                hash_to_group[source_hash] = len(groups) - 1
+            rebuild_index()
+
+        stats: dict[str, dict[str, int]] = {}
+        diff_counts: dict[tuple[str, str], int] = {}
+        diff_samples: list[tuple[str, str, str, str, str]] = []
+        first_tag = str(inputs[0].get("template_tag") or "")
+        base_items = inputs[0].get("papers") or []
+        stats[first_tag] = {"total": len(base_items), "matched": len(base_items), "skipped": 0}
+        for paper in base_items:
+            if not isinstance(paper, dict):
+                raise click.ClickException("Input papers must be objects")
+            db_ops._prepare_paper_matching_fields(paper)
+            add_group(first_tag, paper)
+
+        for bundle in inputs[1:]:
+            template_tag = str(bundle.get("template_tag") or "")
+            items = bundle.get("papers") or []
+            matched = 0
+            skipped = 0
+            for paper in items:
+                if not isinstance(paper, dict):
+                    raise click.ClickException("Input papers must be objects")
+                db_ops._prepare_paper_matching_fields(paper)
+                source_hash = str(paper.get("source_hash") or "")
+                match_idx: int | None = None
+                if source_hash and source_hash in hash_to_group:
+                    match_idx = hash_to_group[source_hash]
+                else:
+                    match_paper, _, _ = db_ops._resolve_paper_by_title_and_meta(
+                        paper, paper_index
+                    )
+                    if match_paper is not None:
+                        match_idx = paper_id_to_group.get(id(match_paper))
+                if match_idx is None:
+                    skipped += 1
+                    continue
+                matched += 1
+                group = groups[match_idx]
+                base_templates = group.get("templates") or {}
+                base_paper = base_templates.get(first_tag)
+                if isinstance(base_paper, dict):
+                    for field in ("source_hash", "paper_title", "publication_date"):
+                        base_value = str(base_paper.get(field) or "")
+                        other_value = str(paper.get(field) or "")
+                        if base_value == other_value:
+                            continue
+                        diff_counts[(template_tag, field)] = diff_counts.get(
+                            (template_tag, field), 0
+                        ) + 1
+                        if len(diff_samples) < 50:
+                            diff_samples.append(
+                                (
+                                    template_tag,
+                                    field,
+                                    str(base_paper.get("paper_title") or ""),
+                                    base_value,
+                                    other_value,
+                                )
+                            )
+                templates = group.setdefault("templates", {})
+                templates[template_tag] = paper
+                order = group.setdefault("template_order", [])
+                if template_tag not in order:
+                    order.append(template_tag)
+            stats[template_tag] = {"total": len(items), "matched": matched, "skipped": skipped}
+
         merged: list[dict[str, Any]] = []
-        for path in input_paths:
-            merged.extend(load_json(Path(path)))
-        write_json(Path(output_path), merged)
-        click.echo(f"Merged {len(input_paths)} files into {output_path}")
+        for group in groups:
+            templates = group.get("templates") or {}
+            order = group.get("template_order") or list(templates.keys())
+            entry: dict[str, Any] = {}
+            for tag in order:
+                paper = templates.get(tag)
+                if not isinstance(paper, dict):
+                    continue
+                for key, value in paper.items():
+                    if key not in entry:
+                        entry[key] = value
+            merged.append(entry)
+
+        output = Path(output_path)
+        write_json(output, merged)
+        _summarize_merge(output, merged, input_count=len(paths))
+
+        stat_table = Table(title="Template Merge Stats")
+        stat_table.add_column("Template")
+        stat_table.add_column("Total", justify="right")
+        stat_table.add_column("Matched", justify="right")
+        stat_table.add_column("Skipped", justify="right")
+        for tag, values in stats.items():
+            stat_table.add_row(
+                tag or "(unknown)",
+                str(values.get("total", 0)),
+                str(values.get("matched", 0)),
+                str(values.get("skipped", 0)),
+            )
+        Console().print(stat_table)
+
+        if diff_counts:
+            diff_table = Table(title="Template Field Diff Summary")
+            diff_table.add_column("Template")
+            diff_table.add_column("Field")
+            diff_table.add_column("Count", justify="right")
+            for (template_tag, field), count in sorted(diff_counts.items()):
+                diff_table.add_row(template_tag or "(unknown)", field, str(count))
+            Console().print(diff_table)
+
+        if diff_samples:
+            sample_table = Table(title="Template Field Diff Samples (up to 50)")
+            sample_table.add_column("Template")
+            sample_table.add_column("Field")
+            sample_table.add_column("Base Title")
+            sample_table.add_column("Base Value")
+            sample_table.add_column("Other Value")
+            for row in diff_samples:
+                sample_table.add_row(*row)
+            Console().print(sample_table)
 
     @db_group.command("render-md")
     @click.option("-i", "--input", "input_path", required=True, help="Input JSON file path")
@@ -843,6 +1252,614 @@ def register_db_commands(db_group: click.Group) -> None:
         rendered = render_papers(papers, out_dir, template, output_language)
         click.echo(f"Rendered {rendered} markdown files")
 
+    @db_group.command("extract")
+    @click.option("--json", "target_json", default=None, help="Target JSON database path")
+    @click.option("--input-json", "input_json", default=None, help="Reference JSON file path")
+    @click.option(
+        "--pdf-root", "pdf_roots", multiple=True, help="PDF root directories for reference (repeatable)"
+    )
+    @click.option(
+        "--md-root", "md_roots", multiple=True, help="Markdown root directories for reference (repeatable)"
+    )
+    @click.option(
+        "--md-translated-root", "md_translated_roots", multiple=True,
+        help="Translated Markdown root directories to extract from (repeatable)"
+    )
+    @click.option(
+        "--md-source-root", "md_source_roots", multiple=True,
+        help="Source Markdown root directories to extract from (repeatable)"
+    )
+    @click.option("--output-json", "output_json", default=None, help="Output JSON file path")
+    @click.option(
+        "--output-md-translated-root",
+        "output_md_translated_root",
+        default=None,
+        help="Output directory for matched translated Markdown",
+    )
+    @click.option(
+        "--output-md-root",
+        "output_md_root",
+        default=None,
+        help="Output directory for matched source Markdown",
+    )
+    @click.option(
+        "-b",
+        "--input-bibtex",
+        "input_bibtex",
+        default=None,
+        help="Reference BibTeX file path",
+    )
+    @click.option("--lang", "lang", default=None, help="Language code for translated Markdown (e.g., zh)")
+    @click.option("--output-csv", "output_csv", default=None, help="Path to export results as CSV")
+    def extract(
+        target_json: str | None,
+        input_json: str | None,
+        pdf_roots: tuple[str, ...],
+        md_roots: tuple[str, ...],
+        md_translated_roots: tuple[str, ...],
+        md_source_roots: tuple[str, ...],
+        output_json: str | None,
+        output_md_translated_root: str | None,
+        output_md_root: str | None,
+        input_bibtex: str | None,
+        lang: str | None,
+        output_csv: str | None,
+    ) -> None:
+        from deepresearch_flow.paper import db_ops
+        from deepresearch_flow.paper.utils import stable_hash
+
+        if input_json and input_bibtex:
+            raise click.ClickException("Use only one of --input-json or --input-bibtex")
+
+        if target_json is None and input_json is not None:
+            target_json = input_json
+
+        has_reference = bool(pdf_roots or md_roots or input_json or input_bibtex)
+        if not has_reference:
+            raise click.ClickException(
+                "Provide at least one reference input: --pdf-root, --md-root, --input-json, or --input-bibtex"
+            )
+        if not target_json and not md_translated_roots and not md_source_roots:
+            raise click.ClickException(
+                "Provide --json and/or --md-translated-root and/or --md-source-root"
+            )
+        if target_json and not output_json:
+            raise click.ClickException("--output-json is required when using --json")
+        if output_json and not target_json:
+            raise click.ClickException("--json is required when using --output-json")
+        if md_translated_roots and not output_md_translated_root:
+            raise click.ClickException(
+                "--output-md-translated-root is required when using --md-translated-root"
+            )
+        if output_md_translated_root and not md_translated_roots:
+            raise click.ClickException(
+                "--md-translated-root is required when using --output-md-translated-root"
+            )
+        if md_source_roots and not output_md_root:
+            raise click.ClickException("--output-md-root is required when using --md-source-root")
+        if output_md_root and not md_source_roots:
+            raise click.ClickException("--md-source-root is required when using --output-md-root")
+        if md_translated_roots and not lang:
+            raise click.ClickException("--lang is required when extracting translated Markdown")
+
+        pdf_root_paths = [Path(path) for path in pdf_roots]
+        md_root_paths = [Path(path) for path in md_roots]
+        translated_root_paths = [Path(path) for path in md_translated_roots]
+        source_root_paths = [Path(path) for path in md_source_roots]
+        reference_json_path = Path(input_json) if input_json else None
+        reference_bibtex_path = Path(input_bibtex) if input_bibtex else None
+
+        reference_papers: list[dict[str, Any]] = []
+        if reference_json_path:
+            if not reference_json_path.is_file():
+                raise click.ClickException(f"Reference JSON not found: {reference_json_path}")
+            reference_papers, _ = load_json_payload(reference_json_path)
+        if reference_bibtex_path:
+            if not reference_bibtex_path.is_file():
+                raise click.ClickException(f"Reference BibTeX not found: {reference_bibtex_path}")
+            if not db_ops.PYBTEX_AVAILABLE:
+                raise click.ClickException("pybtex is required for --input-bibtex support")
+            bib_data = db_ops.parse_file(str(reference_bibtex_path))
+            for key, entry in bib_data.entries.items():
+                title = entry.fields.get("title")
+                if not title:
+                    continue
+                year = entry.fields.get("year") or ""
+                year = str(year) if str(year).isdigit() else ""
+                authors = []
+                for person in entry.persons.get("author", []):
+                    authors.append(str(person))
+                reference_papers.append(
+                    {
+                        "paper_title": str(title),
+                        "paper_authors": authors,
+                        "publication_date": year,
+                        "source_path": f"bibtex:{key}",
+                    }
+                )
+
+        reference_index: dict[str, list[dict[str, Any]]] = {}
+        for paper in reference_papers:
+            if "source_path" not in paper and reference_json_path:
+                paper["source_path"] = str(reference_json_path)
+            db_ops._prepare_paper_matching_fields(paper)
+        if reference_papers:
+            reference_index = db_ops._build_paper_index(reference_papers)
+
+        all_results: list[Any] = []
+
+        if target_json:
+            target_json_path = Path(target_json)
+            if not target_json_path.is_file():
+                raise click.ClickException(f"Target JSON not found: {target_json_path}")
+            papers, payload = load_json_payload(target_json_path)
+
+            results: list[Any] = []
+            matched_indices: set[int]
+            if pdf_root_paths or md_root_paths:
+                results, match_pairs, _, _ = db_ops.compare_datasets_with_pairs(
+                    json_paths_a=[target_json_path],
+                    pdf_roots_b=pdf_root_paths,
+                    md_roots_b=md_root_paths,
+                    bibtex_path=None,
+                    lang=None,
+                    show_progress=True,
+                )
+                matched_indices = {idx_a for idx_a, _, _, _ in match_pairs}
+                all_results.extend(results)
+            else:
+                matched_indices = set(range(len(papers)))
+
+            matched_reference_ids: set[int] = set()
+            if reference_index:
+                def detail_score(paper: dict[str, Any]) -> tuple[int, int]:
+                    non_empty = 0
+                    total_len = 0
+                    for value in paper.values():
+                        if value is None:
+                            continue
+                        if isinstance(value, (list, dict)):
+                            if value:
+                                non_empty += 1
+                                total_len += len(
+                                    json.dumps(value, ensure_ascii=False, sort_keys=True)
+                                )
+                        else:
+                            text = str(value).strip()
+                            if text:
+                                non_empty += 1
+                                total_len += len(text)
+                    return non_empty, total_len
+
+                def resolve_reference_match(
+                    paper: dict[str, Any],
+                ) -> tuple[dict[str, Any] | None, str | None, float]:
+                    match_paper, match_type, match_score = db_ops._resolve_paper_by_title_and_meta(
+                        paper, reference_index
+                    )
+                    if match_paper is not None:
+                        return match_paper, match_type, match_score
+                    year = str(paper.get("_year") or "").strip()
+                    if not year.isdigit():
+                        return None, None, 0.0
+                    authors = paper.get("_authors") or []
+                    author_key = ""
+                    if authors:
+                        author_key = db_ops._normalize_author_key(str(authors[0]))
+                    candidates: list[dict[str, Any]] = []
+                    fallback_type = "year_relaxed"
+                    if author_key:
+                        candidates = reference_index.get(f"authoryear:{year}:{author_key}", [])
+                        if candidates:
+                            fallback_type = "author_year_relaxed"
+                    if not candidates:
+                        candidates = reference_index.get(f"year:{year}", [])
+                    if not candidates:
+                        return None, None, 0.0
+                    title_key = db_ops._normalize_title_key(str(paper.get("paper_title") or ""))
+                    match, score = db_ops._adaptive_similarity_match_papers(title_key, candidates)
+                    if match is None:
+                        return candidates[0], fallback_type, 0.0
+                    return match, fallback_type, score
+
+                base_indices = set(matched_indices)
+                best_matches: dict[int, tuple[int, tuple[int, int], str | None, float]] = {}
+                for idx, paper in enumerate(papers):
+                    if idx not in matched_indices:
+                        continue
+                    db_ops._prepare_paper_matching_fields(paper)
+                    match_paper, match_type, match_score = resolve_reference_match(paper)
+                    if match_paper is None:
+                        continue
+                    ref_id = id(match_paper)
+                    score = detail_score(paper)
+                    current = best_matches.get(ref_id)
+                    if current is None:
+                        best_matches[ref_id] = (idx, score, match_type, match_score)
+                        continue
+                    if score > current[1] or (score == current[1] and match_score > current[3]):
+                        best_matches[ref_id] = (idx, score, match_type, match_score)
+
+                matched_reference_ids = set(best_matches.keys())
+                matched_indices = {idx for idx, *_ in best_matches.values()}
+
+            matched_papers = [paper for idx, paper in enumerate(papers) if idx in matched_indices]
+            deduped_papers: list[Any] = []
+            seen_titles: set[str] = set()
+            for paper in matched_papers:
+                title_key = db_ops._normalize_title_key(str(paper.get("paper_title") or ""))
+                if title_key:
+                    if title_key in seen_titles:
+                        continue
+                    seen_titles.add(title_key)
+                deduped_papers.append(paper)
+            if len(deduped_papers) != len(matched_papers):
+                removed = len(matched_papers) - len(deduped_papers)
+                click.echo(f"Deduplicated {removed} entries by normalized title.")
+            matched_papers = deduped_papers
+            output_path = Path(output_json) if output_json else None
+            if output_path is None:
+                raise click.ClickException("--output-json is required when using --json")
+            output_path.parent.mkdir(parents=True, exist_ok=True)
+            if payload is None:
+                write_json(output_path, matched_papers)
+            else:
+                output_payload = dict(payload)
+                output_payload["papers"] = matched_papers
+                write_json(output_path, output_payload)
+            click.echo(f"Extracted {len(matched_papers)} JSON entries to {output_path}")
+
+            if output_csv and reference_papers:
+                match_meta_by_ref_id = {
+                    ref_id: (idx, match_type, match_score)
+                    for ref_id, (idx, _, match_type, match_score) in best_matches.items()
+                }
+                for ref in reference_papers:
+                    ref_id = id(ref)
+                    ref_title = str(ref.get("paper_title") or "")
+                    ref_hash = stable_hash(str(ref_title or ref.get("source_path") or ""))
+                    ref_path = str(ref.get("source_path") or "")
+                    if ref_id in match_meta_by_ref_id:
+                        idx, match_type, match_score = match_meta_by_ref_id[ref_id]
+                        paper = papers[idx]
+                        paper_hash = str(paper.get("source_hash") or "") or stable_hash(
+                            str(paper.get("paper_title") or "")
+                        )
+                        all_results.append(
+                            db_ops.CompareResult(
+                                side="MATCH",
+                                source_hash=ref_hash,
+                                title=ref_title,
+                                match_status="matched_pair",
+                                match_type=match_type,
+                                match_score=match_score,
+                                source_path=ref_path,
+                                other_source_hash=paper_hash,
+                                other_title=str(paper.get("paper_title") or ""),
+                                other_source_path=str(paper.get("source_path") or ""),
+                                lang=None,
+                            )
+                        )
+                        continue
+                    all_results.append(
+                        db_ops.CompareResult(
+                            side="B",
+                            source_hash=ref_hash,
+                            title=ref_title,
+                            match_status="only_in_B",
+                            match_type=None,
+                            match_score=0.0,
+                            source_path=ref_path,
+                            other_source_hash=None,
+                            other_title=None,
+                            other_source_path=None,
+                            lang=None,
+                        )
+                    )
+
+                for idx in sorted(base_indices - matched_indices):
+                    paper = papers[idx]
+                    paper_title = str(paper.get("paper_title") or "")
+                    paper_hash = str(paper.get("source_hash") or "") or stable_hash(paper_title)
+                    all_results.append(
+                        db_ops.CompareResult(
+                            side="A",
+                            source_hash=paper_hash,
+                            title=paper_title,
+                            match_status="only_in_A",
+                            match_type=None,
+                            match_score=0.0,
+                            source_path=str(paper.get("source_path") or ""),
+                            other_source_hash=None,
+                            other_title=None,
+                            other_source_path=None,
+                            lang=None,
+                        )
+                    )
+
+        copied_count = 0
+        if md_translated_roots:
+            output_root = Path(output_md_translated_root) if output_md_translated_root else None
+            if output_root is None:
+                raise click.ClickException(
+                    "--output-md-translated-root is required when using --md-translated-root"
+                )
+            results, match_pairs, dataset_a, _ = compare_datasets_with_pairs(
+                md_translated_roots_a=translated_root_paths,
+                pdf_roots_b=pdf_root_paths,
+                md_roots_b=md_root_paths,
+                lang=lang,
+                show_progress=True,
+            )
+            matched_indices = {idx_a for idx_a, _, _, _ in match_pairs}
+            copy_iter = tqdm(
+                enumerate(dataset_a.papers),
+                total=len(dataset_a.papers),
+                desc="copy translated",
+                unit="file",
+            )
+            for idx, paper in copy_iter:
+                if idx not in matched_indices:
+                    continue
+                source_path = paper.get("source_path")
+                if not source_path:
+                    continue
+                source = Path(str(source_path))
+                relative = resolve_relative_path(source, translated_root_paths)
+                destination = output_root / relative
+                destination.parent.mkdir(parents=True, exist_ok=True)
+                shutil.copy2(source, destination)
+                copied_count += 1
+            click.echo(
+                f"Copied {copied_count} translated Markdown files to {output_root}"
+            )
+            all_results.extend(results)
+
+        if md_source_roots:
+            output_root = Path(output_md_root) if output_md_root else None
+            if output_root is None:
+                raise click.ClickException("--output-md-root is required when using --md-source-root")
+            results, match_pairs, dataset_a, _ = compare_datasets_with_pairs(
+                md_roots_a=source_root_paths,
+                pdf_roots_b=pdf_root_paths,
+                md_roots_b=md_root_paths,
+                lang=None,
+                show_progress=True,
+            )
+            matched_indices = {idx_a for idx_a, _, _, _ in match_pairs}
+            copied_source = 0
+            copy_iter = tqdm(
+                enumerate(dataset_a.papers),
+                total=len(dataset_a.papers),
+                desc="copy source",
+                unit="file",
+            )
+            for idx, paper in copy_iter:
+                if idx not in matched_indices:
+                    continue
+                source_path = paper.get("source_path")
+                if not source_path:
+                    continue
+                source = Path(str(source_path))
+                relative = resolve_relative_path(source, source_root_paths)
+                destination = output_root / relative
+                destination.parent.mkdir(parents=True, exist_ok=True)
+                shutil.copy2(source, destination)
+                copied_source += 1
+            click.echo(f"Copied {copied_source} source Markdown files to {output_root}")
+            copied_count += copied_source
+            all_results.extend(results)
+
+        if output_csv:
+            output_path = Path(output_csv)
+            export_compare_csv(all_results, output_path)
+            click.echo(f"Results exported to: {output_path}")
+
+    @db_group.command("verify")
+    @click.option("--input-json", "input_json", required=True, help="Input JSON file path")
+    @click.option(
+        "--output-json",
+        "output_json",
+        required=True,
+        help="Output verification report JSON path",
+    )
+    @click.option(
+        "--prompt-template",
+        "prompt_template",
+        default=None,
+        type=click.Choice(list_template_names()),
+        help="Prompt template to load schema (e.g., deep_read)",
+    )
+    @click.option(
+        "-s",
+        "--schema-json",
+        "--schema",
+        "schema_json",
+        default=None,
+        help="Custom schema JSON path",
+    )
+    @click.option(
+        "--ignore-field",
+        "ignore_fields",
+        multiple=True,
+        help="Schema field to ignore when checking empties (repeatable)",
+    )
+    def verify(
+        input_json: str,
+        output_json: str,
+        prompt_template: str | None,
+        schema_json: str | None,
+        ignore_fields: tuple[str, ...],
+    ) -> None:
+        if prompt_template and schema_json:
+            raise click.ClickException("Use only one of --prompt-template or --schema-json")
+        if not prompt_template and not schema_json:
+            raise click.ClickException("Provide --prompt-template or --schema-json")
+
+        input_path = Path(input_json)
+        if not input_path.is_file():
+            raise click.ClickException(f"Input JSON not found: {input_path}")
+
+        papers, payload = load_json_payload(input_path)
+        template_tag = (
+            prompt_template
+            or (payload.get("template_tag") if isinstance(payload, dict) else None)
+            or "custom"
+        )
+
+        try:
+            if schema_json:
+                schema = load_schema(schema_json)
+            else:
+                schema = load_schema_for_template(prompt_template or template_tag)
+        except SchemaError as exc:
+            raise click.ClickException(str(exc)) from exc
+        except ValueError as exc:
+            raise click.ClickException(str(exc)) from exc
+
+        ignore_set = {field.strip() for field in ignore_fields if field.strip()}
+        properties = schema.get("properties", {})
+        schema_fields = sorted(
+            field
+            for field in (set(properties.keys()) | set(schema.get("required", [])))
+            if field not in ignore_set
+        )
+        if not schema_fields:
+            raise click.ClickException("Schema does not define any properties")
+
+        stage_defs = get_stage_definitions(prompt_template or template_tag)
+        field_stage_map: dict[str, str] = {}
+        for stage_def in stage_defs:
+            for field in stage_def.fields:
+                if field in ignore_set:
+                    continue
+                field_stage_map.setdefault(field, stage_def.name)
+
+        report_items: list[dict[str, Any]] = []
+        for paper in papers:
+            if not isinstance(paper, dict):
+                continue
+            missing_fields = [
+                field
+                for field in schema_fields
+                if field not in paper or is_empty_value(paper.get(field))
+            ]
+            if not missing_fields:
+                continue
+            item: dict[str, Any] = {
+                "source_path": str(paper.get("source_path") or ""),
+                "paper_title": str(paper.get("paper_title") or ""),
+                "missing_fields": missing_fields,
+            }
+            if field_stage_map and all(field in field_stage_map for field in missing_fields):
+                item["retry_stages"] = sorted(
+                    {field_stage_map[field] for field in missing_fields}
+                )
+            report_items.append(item)
+
+        report_payload = {
+            "template_tag": template_tag,
+            "schema_fields": schema_fields,
+            "items": report_items,
+        }
+
+        output_path = Path(output_json)
+        output_path.parent.mkdir(parents=True, exist_ok=True)
+        write_json(output_path, report_payload)
+
+        console = Console()
+        total_missing = sum(len(item["missing_fields"]) for item in report_items)
+        summary_table = Table(title="db verify summary")
+        summary_table.add_column("Metric", style="cyan")
+        summary_table.add_column("Value", style="white", overflow="fold")
+        summary_table.add_row("Input", str(input_path))
+        summary_table.add_row("Template", template_tag)
+        summary_table.add_row("Items", str(len(papers)))
+        summary_table.add_row("Items with missing fields", str(len(report_items)))
+        summary_table.add_row("Total missing fields", str(total_missing))
+        if ignore_set:
+            summary_table.add_row("Ignored fields", ", ".join(sorted(ignore_set)))
+        summary_table.add_row("Output", str(output_path))
+        console.print(summary_table)
+
+        if report_items:
+            field_counts: dict[str, int] = {field: 0 for field in schema_fields}
+            for item in report_items:
+                for field in item["missing_fields"]:
+                    field_counts[field] = field_counts.get(field, 0) + 1
+
+            count_table = Table(title="Missing field counts")
+            count_table.add_column("Field", style="cyan")
+            count_table.add_column("Missing", style="yellow", justify="right")
+            for field, count in sorted(field_counts.items(), key=lambda x: (-x[1], x[0])):
+                if count:
+                    count_table.add_row(field, str(count))
+            console.print(count_table)
+
+            detail_table = Table(title="Missing field details")
+            detail_table.add_column("#", style="dim", justify="right")
+            detail_table.add_column("Title", style="white", overflow="fold")
+            detail_table.add_column("Source Path", style="cyan", overflow="fold")
+            detail_table.add_column("Missing Fields", style="yellow", overflow="fold")
+            detail_table.add_column("Retry Stages", style="green", overflow="fold")
+            for idx, item in enumerate(report_items, start=1):
+                retry_stages = item.get("retry_stages") or []
+                detail_table.add_row(
+                    str(idx),
+                    item.get("paper_title") or "",
+                    item.get("source_path") or "",
+                    ", ".join(item.get("missing_fields", [])),
+                    ", ".join(retry_stages),
+                )
+            console.print(detail_table)
+        else:
+            console.print(Panel("[green]No missing fields detected.[/green]", expand=False))
+
+    @db_group.command("transfer-pdfs")
+    @click.option("--input-list", "input_list", required=True, help="Text file containing PDF paths")
+    @click.option("--output-dir", "output_dir", required=True, help="Output directory")
+    @click.option("--move", "move_files", is_flag=True, help="Move PDFs instead of copying")
+    @click.option("--copy", "copy_files", is_flag=True, help="Copy PDFs instead of moving")
+    def transfer_pdfs(
+        input_list: str,
+        output_dir: str,
+        move_files: bool,
+        copy_files: bool,
+    ) -> None:
+        if move_files == copy_files:
+            raise click.ClickException("Specify exactly one of --move or --copy")
+
+        list_path = Path(input_list)
+        if not list_path.is_file():
+            raise click.ClickException(f"Input list not found: {list_path}")
+
+        destination_root = Path(output_dir)
+        destination_root.mkdir(parents=True, exist_ok=True)
+
+        entries = [line.strip() for line in list_path.read_text(encoding="utf-8").splitlines()]
+        entries = [line for line in entries if line]
+
+        processed = 0
+        missing = 0
+        transfer_iter = tqdm(entries, total=len(entries), desc="transfer pdfs", unit="file")
+        for raw in transfer_iter:
+            source = Path(raw).expanduser()
+            if not source.is_file():
+                missing += 1
+                continue
+            destination = destination_root / source.name
+            destination.parent.mkdir(parents=True, exist_ok=True)
+            if move_files:
+                shutil.move(str(source), str(destination))
+            else:
+                shutil.copy2(source, destination)
+            processed += 1
+
+        action = "Moved" if move_files else "Copied"
+        click.echo(f"{action} {processed} PDFs to {destination_root}")
+        if missing:
+            click.echo(f"Skipped {missing} missing paths")
+
     @db_group.command("compare")
     @click.option(
         "-ia", "--input-a", "input_paths_a", multiple=True, help="Input JSON files for side A (repeatable)"
@@ -875,6 +1892,12 @@ def register_db_commands(db_group: click.Group) -> None:
     @click.option(
         "--output-csv", "output_csv", default=None, help="Path to export results as CSV"
     )
+    @click.option(
+        "--output-only-in-b",
+        "output_only_in_b",
+        default=None,
+        help="Path to export only-in-B source paths as a newline list",
+    )
     @click.option(
         "--sample-limit", "sample_limit", default=5, type=int, show_default=True,
         help="Number of sample items to show in terminal output"
@@ -891,12 +1914,12 @@ def register_db_commands(db_group: click.Group) -> None:
         bibtex_path: str | None,
         lang: str | None,
         output_csv: str | None,
+        output_only_in_b: str | None,
         sample_limit: int,
     ) -> None:
         """Compare two datasets and report matches and differences."""
         from deepresearch_flow.paper.db_ops import compare_datasets
-        import csv
-        
+
         # Validate that at least one input is provided for each side
         has_input_a = bool(input_paths_a or pdf_roots_a or md_roots_a or md_translated_roots_a)
         has_input_b = bool(input_paths_b or pdf_roots_b or md_roots_b or md_translated_roots_b)
@@ -925,6 +1948,7 @@ def register_db_commands(db_group: click.Group) -> None:
                 md_translated_roots_b=[Path(p) for p in md_translated_roots_b],
                 bibtex_path=Path(bibtex_path) if bibtex_path else None,
                 lang=lang,
+                show_progress=True,
             )
         except ValueError as exc:
             raise click.ClickException(str(exc)) from exc
@@ -998,31 +2022,14 @@ def register_db_commands(db_group: click.Group) -> None:
         # Export to CSV if requested
         if output_csv:
             output_path = Path(output_csv)
-            output_path.parent.mkdir(parents=True, exist_ok=True)
-            
-            with open(output_path, "w", newline="", encoding="utf-8") as f:
-                writer = csv.writer(f)
-                writer.writerow([
-                    "Side", "Source Hash", "Title", "Match Status", "Match Type",
-                    "Match Score", "Source Path", "Other Source Hash", "Other Title",
-                    "Other Source Path", "Lang"
-                ])
-                for r in results:
-                    writer.writerow([
-                        r.side,
-                        r.source_hash,
-                        r.title,
-                        r.match_status,
-                        r.match_type or "",
-                        f"{r.match_score:.4f}",
-                        r.source_path or "",
-                        r.other_source_hash or "",
-                        r.other_title or "",
-                        r.other_source_path or "",
-                        r.lang or "",
-                    ])
-            
+            export_compare_csv(results, output_path)
             console.print(f"\n[green]Results exported to: {output_path}[/green]")
-        
+        if output_only_in_b:
+            output_path = Path(output_only_in_b)
+            count = export_only_in_b_paths(results, output_path)
+            console.print(
+                f"\n[green]Only-in-B list exported ({count} items): {output_path}[/green]"
+            )
+
         # Print final counts
         console.print(f"\nTotal results: {len(results)}")